Clinical Characteristics and Outcomes of Central Nervous System Tumors Harboring NTRK Gene Fusions

NTRK1, NTRK2, and NTRK3 genes code for the tropomyosin receptor kinase (TRK) family of receptors TRKA, TRKB, and TRKC, which are neurotrophic tyrosine receptor kinase (NTRK) proteins (1, 2). These receptors are expressed in neuronal tissue and play an essential role in the normal development and function of the nervous system (1, 3). NTRK gene fusions have been reported in a variety of pediatric and adult tumors and occur when the 3′ region of the NTRK gene encoding the tyrosine kinase domain is joined in-frame with the 5′ end of a fusion partner gene, either by intra- or inter-chromosomal rearrangement (4). The resulting fusion oncogene leads to the expression of a chimeric protein that retains the tyrosine kinase domain, is constitutively active, and drives downstream signaling (4).

NTRK gene fusions occur in up to 1% of all solid tumors and 2% of adult primary central nervous system (CNS) tumors (2, 5–7). In the pediatric population, NTRK gene fusions have been observed in up to 5.3% of high-grade gliomas (HGG) and 2.5% of low-grade gliomas (LGG; refs. 7–9). Recently, a subset of gliomas with NTRK gene fusions has been grouped under the tumor type infant-type hemispheric glioma in the 2021 World Health Organization Classification of Tumors of the CNS (10).

There are limited data on CNS tumors with NTRK fusion in the literature, mostly presented in case reports and small case series or included in larger studies focused on molecular characterization of pediatric CNS tumors (11–22). The largest cohort included 33 patients enrolled on two clinical trials and treated with larotrectinib, a selective TRK inhibitor (TRKi; ref. 23). Larotrectinib and entrectinib have been FDA- and European Medicines Agency–approved in a histology agnostic fashion for patients with NTRK fusion–positive solid tumors and are slowly starting to impact the management of CNS tumors with NTRK fusion: (24). However, the natural history and outcomes of patients with NTRK fusion–positive CNS tumors are not well described. A better understanding of these rare tumors will help interpret the efficacy and limitations associated with these new targeted therapy approaches. Herein, we report the characteristics and outcomes of a large international cohort of pediatric and adult patients with CNS tumors harboring NTRK fusions.

This is an international multicenter retrospective cohort study of patients with CNS tumor and NTRK fusion. All patients diagnosed between 2000 and 2021 with a confirmed TRK fusion were eligible. To identify patients, an invitation email was sent to oncologists and neuro-oncologists from international sites. Patients identified to have an NTRK gene fusion in the Children’s Brain Tumor Network (CBTN) database were also included. The study was conducted in accordance with the Declaration of Helsinki and was approved by an Institutional Review Board-Centre Hospitalier Universitaire (IRB-CHU Sainte-Justine). Written informed consent was waived by the IRB given the retrospective nature of the study and by the fact that data were coded and protected health information was removed.

After institutional approval, centers received a standardized case report form which included patient’s demographics, pathology characteristics, tumor location, treatments, and outcomes. Treatment regimens were collected and categorized as surgery, chemotherapy, radiotherapy, TRKi therapy, and other. For analysis purposes, “nontargeted therapy” included chemotherapy, radiotherapy, and other and excluded surgery only and TRKi therapy. For CBTN subjects, these data were extracted from the CBTN database with additional queries to treating sites as necessary. When possible, response was assessed by the local investigator and categorized as complete response (CR), partial response (PR), minor response (MR), stable disease, and progressive disease and was considered NE (not evaluable) if a gross total or near total resection was done prior to treatment (Supplementary Data S1). Centers were encouraged to use the response assessment in neuro-oncology or response assessment in pediatric neuro-oncology criteria (25, 26). For patients enrolled in clinical trials, responses were not extracted to avoid interference and confidentiality breach. Clinically significant response was defined as CR, PR, or MR.

Descriptive statistics were reported as counts and percentages for categorical variables and median and range for continuous variables. Outcomes, including progression free survival (PFS) and overall survival (OS) were presented using Kaplan–Meier survival curves. If relevant, groups were compared using the log-rank test. The growth modulation index (GMI) was calculated based on the method described by Von Hoff and colleagues (27). The GMI is the ratio of the PFS of a treatment compared with the time to progression (TTP) of the previous treatment (PFS_treatment/TTP_{previous therapy}). Patients are censored at progression or the last follow-up or the most recent report. It is usually accepted that a GMI ≥1.3 suggests a significant treatment benefit with a PFS which increases by 30% or more in the second-line therapy. The GMI has been used in several precision medicine trials, including recent studies (28–30). In our study, patients were included in the GMI analysis if they received a prior line of treatment (chemotherapy or radiotherapy) followed by progression and initiation of TRKi therapy. Patients with a combination of TRKi therapy and other systemic therapy or radiation were excluded. The Kaplan–Meier estimate was obtained using the GMI calculated as the ratio of the TTP/censoring with TRKi therapy to the TTP with previous therapy (x-axis) and the PFS rate on TRKi therapy (y-axis). At GMI = 0, all patients treated with TRKis were at risk of progression and the survival probability was 1. The Kaplan–Meier estimate was used to describe the probability of PFS on TRKi therapy past a specific value of GMI (1.3). The higher the probability of GMI ≥1.3, the higher the benefit of TRKi therapy compared with previous treatment. All statistical tests were two-sided and conducted at the 0.05 significance level. Statistical analyses were performed using Statistical Analysis System version 9.4.

Patient data are not publicly available to protect patient privacy but are available upon reasonable request to the corresponding author (S. Perreault, [email protected]). Provided data will be deidentified.

A total of 170 investigators from 41 countries were contacted by email. In total, 129 patients from 46 centers (21 countries) were identified. Ten patients with a reported NTRK alteration were excluded as it was not possible to confirm whether their tumor had an NTRK fusion; thus, the cohort included 119 patients. Thirty-one patients (26.1%) had previously been reported in the literature as part of a clinical trial, case series, or case report (11–22) Supplementary Fig. S1.

The median follow-up of the entire cohort was 38.5 months (range 0.03–229.3 months; Table 1). The median age at time of diagnosis was 4.5 years (range 0–78.4 years), and almost half of all patients were infants of less than 3 years of age (n = 53/117, 45.3%). The median age of the adult cohort was 50 years (range 18–78.4 years).

The most frequent location for CNS tumors with an NTRK fusion was hemispheric (63.8%); 13 patients (13%) were found to have metastatic disease (Table 1). The majority were reported to have a histology consistent with a diagnosis of HGG (57.1%), followed by LGG (27.7%), embryonal tumors (4.2%), and others (10.9%; Table 1). All 17 adult patients had a diagnosis of HGG. In children, LGG and HGG had a different distribution according to age, with HGG predominantly found in patients under 3 years of age (72%, P = 0.0035; Table 1). Most patients had an NTRK2 fusion (50.4%; Fig. 1). The specific NTRK gene (NTRK1, NTRK2, or NTRK3) was not associated with a specific histology (Supplementary Table S1). Sequencing was done at primary site using next-generation sequencing for NTRK fusion and other alterations (Supplementary Tables S2 and S3).

For initial treatment, 40 patients (33.6%) had surgery only and then were observed without further treatment, 72 patients (60.5%) received either chemotherapy, radiation, or other systemic therapy excluding TRKis, and 7 patients (5.9%) were treated upfront with TRKis (Table 2).

When assessing treatments received during the observation period, 28 patients (23.5%) underwent surgery only without further treatment, most of these patients were pediatric patients with LGG (60.7%), and 17.9% were children with HGG (P < 0.0001). A total of 43 patients (36.1%) received one line of treatment besides surgery, and 48 patients (40.3%) received three or more lines of treatment (Table 2). A total of 51 patients (42.9%) were treated with TRKis at some point during their follow-up, including 39 patients (76.5%) with larotrectinib, 3 patients (5.9%) with entrectinib, and 9 patients (17.6%) with other or nonspecified TRKis (Table 2). The first patient who received a TRKi was treated in 2015. Four patients received two different TRKis, and one patient was reported to have received three different TRKis. Sixteen patients (13.4%) were enrolled on a therapeutic clinical trial of a TRKi; eight with larotrectinib (two of whom subsequently received selitrectinib), three with entrectinib, and five with an “unspecified” TRKi.

Nontargeted therapy was given 99 times in 65 patients, and response was assessed for 63 of these 99 lines of treatment. Of these 63 evaluable nontargeted therapy regimens, 24 (38.1%) resulted in an objective response (10 CR, 13 PR, and 1 MR; Supplementary Table S4). For 35 evaluable nontargeted regimens, the pediatric HGG objective response was 42.9% (six CR, eight PR, and one MR). The infant (<3 years) HGG objective response was 45.2% (six CR, seven PR, and one MR) among 31 evaluable nontargeted regimens. The older children/adolescent HGG objective response was 25% (one PR) among four evaluable nontargeted regimens.

Response to TRKi therapy was evaluable for 33 separate instances of treatment overall and for 21 instances of treatment for pediatric patients. The response rate was 42.4% overall (five CR, seven PR, and two MR) and 61.9% for the pediatric patients (four CR, seven PR, and two MR; Supplementary Table S4). When restricted to patients with HGG, response to TRKi therapy was evaluable for 22 separate treatment instances, with an overall response rate of 45.5% (five CR, four PR, and one MR) and a 90% response rate (four CR, four PR, and one MR) in children.

Larotrectinib was given 23 times in 22 pediatric patients, and response was assessed following 16 of these 23 lines of treatment. Of these 16 evaluable larotrectinib regimens, 11 (68.8%) demonstrated objective response (four CR, five PR, and two MR). When restricted to patients with pediatric HGG, response to larotrectinib was evaluable for eight patients, with an overall response rate of 100% (four CR, three PR, and one MR; Supplementary Table S4).

The median follow-up was 38.5 months (range: 0.03–229.3). At the last follow-up, 88 patients (74%) were alive (Supplementary Table S5). The OS and PFS were analyzed for the entire cohort and according to patients’ characteristics (Figs. 2 and 3; Supplementary Fig. S2). The median OS of the entire cohort was 185.5 months [95% confidence interval (CI), 99.5–229.3; Fig. 2A], and the median PFS was 25.5 months (95% CI, 15.5–40.9; Fig. 3A). Gross total resection and near total resection were not associated with better OS (P = 0.45) or PFS (P = 0.40) when compared with partial resection or biopsy.

Pediatric patients had a better prognosis with a median OS of 185.5 months (95% CI, 185.5–229.3) compared with 24.8 months in adults (95% CI, 17.1–99.5; P < 0.0001; Fig. 2B). However, the median PFS between pediatric and adult patients was not significantly different (25.8 vs. 11.1 months, respectively P = 0.22; Fig. 3B). Patients with LGG also had a better prognosis with a median OS that was not reached compared with 99.5 months for HGG (95% CI, 57.9–229.3) and 38.5 (95% CI, 3.5–NE) months for embryonal tumors (P = 0.0012; Fig. 2C). This difference in OS by histology was also significant in the pediatric cohort (P = 0.021). Pediatric patients with HGG had a better outcome when compared with adults with HGG. The median OS with HGG was 185.5 months (95% CI, 65.3–NE) for children compared with 24.8 (95% CI, 17.1–NE) months for adults (P = 0.0035). There was no difference in PFS between histologic subtype (P = 0.30; Fig. 3C). Six patients with HGG underwent surgery only, and two remained alive at the last follow-up (range 28.7–43.7 months). NTRK gene fusion type (1, 2, or 3) was not associated with a difference in OS (P = 0.18) or (Fig. 2D; Supplementary Fig. S2A). We observed a tendency for an improved PFS in pediatric patients with NTRK1 or NTRK2 in the present cohort (P = 0.0526 for the comparison NTRK1 vs. NTRK3 and P = 0.0214 for the comparison NTRK2 vs. NTRK3; Fig. 3D and Supplementary Fig. S2B). Patients with CDKN2A/B alteration had a median OS of 57.9 months compared with 229.3 months for patients without alteration (P = 0.053; Fig. 2E). Most patients with CDKN2A/B alteration had a diagnosis of HGG (16/18%–88.9%). There was no significant difference in PFS between patients with and without CDKN2A/B alteration (Fig. 3E and F). Only one patient’s tumor classified as LGG had a CDKN2A/B deletion. This infant had a cerebellar lesion that underwent two resections without systemic therapy or radiotherapy and was still alive at a follow-up of 57 months. Finally, no difference was observed in OS of children with HGG who received radiotherapy or did not (P = 0.695; Fig. 2F).

The GMI was calculated in 31 patients (20 pediatric and 11 adults). Twenty patients were excluded from the analysis (11 received TRKi therapy plus another treatment, 5 did not received a treatment prior to a TRKi, 2 had no available data, and 2 were treated with a TRKi but without prior progression). The median GMI was 1.1 (range 0.09–11.51), and 15/31 (48.4%) had a GMI ≥1.3. For the pediatric population, the median GMI was 1.36 (range 0.15–11.51), and 10/20 (50.0%) had a GMI ≥1.3. Eight patients (40%) remained on TRKi therapy at the last reported time point. The average time on TRKi therapy was 26 months, and no patient without progression discontinued treatment. For the pediatric subgroup with HGG (N = 12), the median GMI was 0.6 (range 0.15–7.58), and 4/12 (33.3%) had a GMI ≥1.3.

The Kaplan–Meier estimate for the probability of having a GMI ≥1.3 for TRKi was 0.62 (95% CI, 0.36–0.80) in pediatric patients [this probability was 0.44 (95% CI, 0.15–0.70) in the pediatric subgroup with HGG (N = 12)] and 0.45 (95% CI, 0.17–0.71) in adult patients (N = 11). The PFS for pediatric patients treated with TRKi therapy was therefore better compared with previous therapy (Fig. 4A). However, this difference was not observed in the pediatric HGG subgroup nor in adult patients (Fig. 4B).

Specifically for larotrectinib, the GMI was calculated as a ratio of PFS with larotrectinib to the TTP of the prior line of therapy in 23 patients (14 pediatric and 9 adults). The median GMI was 1.66 (range 0.09–11.51), and 14/23 (60.9%) had a GMI ≥1.3. For the pediatric population, the median GMI was 2.11 (range 0.15–11.51), and 10/14 (71.4%) had a GMI ≥1.3. Specific median GMI to compare pediatric LGG (N = 2) with HGG (N = 8) was not possible due to the small number of patients in the pediatric subgroup with LGG available for this analysis. For the pediatric subgroup with HGG (N = 8), the median GMI was 1.3 (range 0.15–7.58), and 4/8 (50.0%) had a GMI ≥1.3.

Using the Kaplan–Meier estimate, the probability of having a GMI ≥1.3 for larotrectinib was 0.85 (95% CI, 0.52–0.96) in pediatric patients [this probability was 0.73 (95% CI, 0.28–0.93) in the pediatric subgroup with HGG (N = 8)] and 0.44 (95% CI, 0.14–0.72) in adult patients (N = 9; Fig. 4C and D). The PFS with larotrectinib was superior when compared with previous therapies for pediatric patients (including the pediatric HGG subgroup), but this difference was not observed in adult patients (Fig. 4C and D).

To our knowledge, in this study, we present the largest cohort of patients with CNS tumors and confirmed NTRK fusions. We show that pediatric patients and those with tumors classified as LGG have better OS. Furthermore, there is also evidence that TRKis can provide better overall response and PFS when compared with previous therapies, especially in children.

The outcome of pediatric patients is significantly better when compared with adults, with a median OS of more than 15 years compared with 2 years for adults. However, the pediatric cohort included a mixture of histology, whereas adult cohort included only HGG. Amongst HGG, adults had a worse outcome. Our observations are in line with what has been reported in the SCOUT/NAVIGATE trials. When treated with larotrectinib, no adult reached a PR compared with 40% of children (23, 31). This difference in outcome will be important to account for in ongoing and future clinical trials. Within the present study, additional risk factors that could explain this difference were not identified. Additional molecular alterations that would differentiate pediatric from adult tumors were not reported in this dataset, as central testing was not performed. However, NTRK fusions may be late events in the pathophysiology of adult CNS tumors or serve as one of multiple oncogenic mutations in adult tumors, in contrast to children in whom NTRK fusions are thought to act as the primary driver mutation. To answer this question, next-generation sequencing data and DNA methylation profiles will need to be collected and correlated with outcomes.

Whereas we acknowledge that glioma grading can be challenging, especially in young children, our data show that grading based on histology seems to remain an important predictor of outcome in CNS tumors with NTRK fusions. This concept will also need to be validated by central review of histology, molecular features, and methylation analysis.

We observed a tendency for an improved PFS in pediatric patients with NTRK1 in the present cohort. This observation has not been reported in other cancers and might not be clinically significant. However, this observation should be explored in future cohort studies and clinical trials. Our study showed no difference in fusion partner but also will need prospective follow-up to confirm that the fusion partner is not of significance.

We report that almost a quarter of patients, including two with HGG, underwent surgery only and were long-term survivors. Based on this observation, it is possible that select patients who undergo a gross total resection could be cautiously observed before initiating systemic therapy, including those with pediatric HGG, given their outstanding OS (median OS 185.5 months). In our study, we did not observe a better outcome in children with HGG who received radiotherapy compared with those who did not. The decision to use radiotherapy should therefore be balanced against the potential side effects, especially in young children. In addition, given the observed GMI and excellent response rates to TRKis, it may be reasonable to consider limited resection in specific cases of pediatric HGG and offer a TRKi as an initial treatment. We did not observe a significant difference between TRKi therapy and systemic therapy for pediatric HGG, but given the toxicity profile, the use of TRKis is a new interesting avenue. The response rate and GMI of HGG treated with larotrectinib were particularly high and significantly more efficacious when compared with systemic treatment. Upfront therapy using TRKis is under investigation (NCT04655404).

Not surprisingly, more than half of pediatric patients with LGG underwent surgery only. LGG with NTRK fusions do not seem to be at higher risk of requiring additional lines of treatment when compared with historic LGG data. However, given the fact that some of these lesions might be difficult to resect, surgery with high risk of neurologic deficits should be avoided. In these specific cases of LGG, standard chemotherapy should be considered and TRKis might offer an interesting alternative especially in the context of recurrent disease. Given the small number of LGG cases available for response analysis and GMI analysis, no conclusion on efficacy of TRKi therapy compared with chemotherapy can be drawn.

Assessing the efficacy of treatment in a retrospective cohort is challenging. The evaluation of response was not centralized but rather based on local investigator evaluation. Clinical practice does not always follow formal response assessment in neuro-oncology/response assessment in pediatric neuro-oncology response criteria. Regardless, we observed that pediatric patients had a better response rate to larotrectinib when compared with nontargeted therapy, suggesting that this treatment approach might yield a clinical benefit. The efficacy of other TRKis could not be evaluated given the small number of patients.

Comparing response rates at different time points is also a significant limitation. A patient facing multiple relapses may be less likely to respond to treatment. The GMI is an innovative measure which is useful for relapsed and refractory cancers. Its clinical application has been accepted to evaluate the efficacy of treatment using an intrapatient control. In a given patient who had progression, an effective therapy should increase the next TTP. In our study, we demonstrated a GMI of 2.1 for the pediatric patients treated with larotrectinib, which is substantially higher than the 1.33 GMI cut-off associated with a probable efficacy. Although arbitrary, this threshold of 33% improvement seems appropriate, as PFS tends to decrease with subsequent lines of systemic therapy in other solid tumors (32). A number of the patients included remained on treatment as of the data cut-off, suggesting that the median GMI may increase further. We suggest that the GMI could be integrated in ongoing studies as a secondary objective to evaluate the efficacy of TRKis. One limitation of the GMI is that it selects against patients who succumb after the first line of therapy, because these critical patients are removed from the analysis. Another limitation is the small sample size of patients without HGG evaluable for analysis, thus limiting our ability to generalize these observations to other histologies.

Whereas we broadly reached out to providers by email to identify potential patients, it is likely that we gathered more responses from centers with which we had previously established collaborations through other clinical studies. Both the fact that two of the principal authors have been involved in clinical trials involving larotrectinib and the young age of this cohort (larotrectinib is the only FDA-approved TRKi for patients below 12 years of age) might explain why we collected more patients treated with larotrectinib as compared with other TRKis. Another potential limitation of this study is that only patients locally identified to have NTRK fusions were included. This may bias toward patients with worse outcomes, as testing may have been performed more frequently in patients with difficult to treat or relapsed disease.

Finally, despite the fact we reported the largest cohort of patients with CNS tumors and NTRK fusion, the study includes a small number of patients with embryonal tumors and even LGG. We are planning to continue data collection and increase the number of patients, but there is an urgent need for future prospective clinical trials addressing the current limitations of our data analysis.

In summary, we describe a large cohort of patients with CNS tumors and NTRK fusion. We identified that young age and low-grade histology are associated with improved outcomes. TRKi therapy seems to improve tumor control in a subset of patients, most notably for pediatric HGG. Additional prospective study and clinical trials are needed to improve management of patients with CNS tumors and NTRK fusion. Standard treatments such as chemotherapy and radiotherapy could be compared with upfront treatment with TRKis. Minimally invasive surgery followed by treatment with TRKis and second-look surgery could also be investigated within a clinical trial.

M.J. Fisher reports grants and other support from Alexion/AstraZeneca, other support from SpringWorks and Day 1, and grants from Array Biopharma/Pfizer and Exelixis outside the submitted work. L. Lemelle reports grants from Bayer Inc. and SFCE—Imagine for Margo during the conduct of the study. C. Kramm reports grants, personal fees, and nonfinancial support from Merck and Deutsche Kinderkrebsstiftung; personal fees and nonfinancial support from Rover/Blueprint and Novartis; personal fees from Novogene and Boehringer Ingelheim; and nonfinancial support from Oncoscience during the conduct of the study. S.M. Pfister reports nonfinancial support from PMC and Epignostix GmbH and personal fees from BioSkryb outside the submitted work; in addition, S.M. Pfister shares a patent with D.T.W. Jones, David Capper, Andreas von Deimling, Martin Sill, and Volker Hovestadt for DKFZ, University Hospital Heidelberg DNAMethylation based method for classifying tumor species (European Patent 16 710700.2) issued and has unpaid membership in GPOH, SIOP, DGNN, and DKG. D.T.W. Jones reports personal fees from Day 1 Biopharmaceuticals and other support from Heidelberg Epignostix GmbH outside the submitted work. D. Orbach reports grants from Bayer Inc. and SFCE—Imagine for Margo during the conduct of the study. G. Pierron reports grants from Bayer Inc. and SFCE—Imagine for Margo charity during the conduct of the study. A. Drilon reports personal fees from 14ner/Elevation Oncology, Amgen, Anheart Therapeutics, AbbVie, ArcherDX, AstraZeneca, Bayer, Beigene, BergenBio, Blueprint Medicines, Boundless Bio, Bristol Myers Squibb, Chugai Pharmaceutical, EcoR1, EMD Serono, Entos, Exelixis, Helsinn, Hengrui Therapeutics, Ignyta/Genentech/Roche, Innocare, Janssen, Loxo/Lilly, Merus, Monopteros, MonteRosa, Novartis, Nuvalent, Pfizer, Prelude, Regeneron, Repare RX, Takeda/Ariad/Millenium, Treeline Bio, TP Therapeutics, Tyra Biosciences, Verastem, and Zymeworks and other support from mBrace, Treeline, Boehringer Ingelheim, Merck, and Puma during the conduct of the study as well as other support from Foundation Medicine, Teva, Taiho, GlaxoSmithKline, and PharmaMar and personal fees from Answers in CME, Applied Pharmaceutical Science, Inc., AXIS, Clinical Care Options, Doc Congress, EPG Health, Harborside Nexus, I3 Health, Imedex, Liberum, Medendi, Medscape, Med Learning, MedTalks, MJH Life Sciences, MORE Health, Ology, OncLive, Paradigm, Peerview Institute, PeerVoice, Physicians Education Resources, Projects in Knowledge, Resources, Remedica Ltd., Research to Practice, RV More, Springer Healthcare, Targeted Oncology, TouchIME, WebMD, Wolters Kluwer, and UpToDate outside the submitted work; in addition, A. Drilon has a patent for Selpercatinib-Osimertinib (US 18/041617, pending) pending. E.L. Diamond reports personal fees from Opna Bio, nonfinancial support from Pfizer Inc., and personal fees from Day 1 Biopharmaceuticals and SpringWorks Therapeutics outside the submitted work. G. Harada reports personal fees from Bristol Myers Squibb, MSD, Lilly, Takeda, Pfizer, AstraZeneca, Daiichi, Sanofi, Novartis, and J&J outside the submitted work. M. Zapotocky reports personal fees from AstraZeneca outside the submitted work. A.G. Weil reports personal fees from Monteris Medical Inc. and Synergia SA outside the submitted work. H. Coltin reports personal fees from Alexion Pharma Canada Corp. outside the submitted work. R. Hammad reports other support from Vertex outside the submitted work. J.R. Hansford reports personal fees from Servier International, Alexion Pharmaceuticals, and Bayer Pharmaceuticals outside the submitted work. N.G. Gottardo reports other support from Bayer outside the submitted work. H. Goto reports personal fees from Bayer during the conduct of the study. D.S. Ziegler reports personal fees from Bayer during the conduct of the study as well as grants and personal fees from Accendatech and personal fees from Novartis, Day 1, FivePhusion, Amgen, Alexion, Norgine, and Medison Pharma outside the submitted work. H. Lindsay reports personal fees from Day 1 Bio outside the submitted work. Y.-L. Liu reports grants and personal fees from Bayer; grants from Turning Point Therapeutics Inc., a subsidiary of Bristol Myers Squibb Co.; and personal fees from AstraZeneca, Roche, Novartis, JW Diagnostics, and CancerFree Biotech outside the submitted work. A.T. Franson reports personal fees from Bayer outside the submitted work. E. Hwang reports personal fees from Day 1 Therapeutics outside the submitted work. S. Cheng reports other support from Bayer outside the submitted work. L.M. Hoffman reports personal fees from Day 1 Biopharmaceuticals and Bayer outside the submitted work. A. Lassaletta reports other support from Alexion and Servier and personal fees from Pfizer outside the submitted work. N.J. Robison reports grants from Pfizer outside the submitted work. S. Abbou reports other support from Fore Pharmaceutical and Lilly outside the submitted work; in addition, S. Abbou has a patent for EP3613436B1 issued. P. Berlanga reports other support from Bayer and EUSA Pharma/Recordati outside the submitted work. C.L. Moertel reports other support from OX2 Therapeutics and personal fees from SpringWorks Therapeutics and Alexion Pharmaceuticals outside the submitted work; in addition, C.L. Moertel has a patent for Patent No. US 9364505 B2 issued and a patent for International Patent No. WO 2020/150149 A1 issued. E.D. Razis reports personal fees from Bayer, AstraZeneca, Servier, Novartis, Gilead, Pfizer, Genesis Pharm, and Bristol Myers Squibb outside the submitted work. F. Doz reports personal fees from Roche, Ipsen, Servier, Novartis Pharma S.A.S.; grants from Onxeo Valerio; and nonfinancial support from Synth-Innove outside the submitted work. T.W. Laetsch reports grants and personal fees from Bayer and Jazz Pharmaceuticals; personal fees from GSK, AI Therapeutics, Advanced Microbubbles, and ITM Oncologics; and grants from Pfizer outside the submitted work. S. Perreault reports grants from Roche and Bayer during the conduct of the study as well as grants and personal fees from Alexion, grants from Novartis and SpringWorks Therapeutics, and personal fees from Bayer and Esai outside the submitted work. No disclosures were reported by the other authors.

A.-A. Lamoureux: Conceptualization, data curation. M.J. Fisher: Conceptualization, data curation. L. Lemelle: Conceptualization, data curation. E. Pfaff: Conceptualization, data curation. P. Amir-Yazdani: Conceptualization, data curation. C. Kramm: Conceptualization, data curation. B. De Wilde: Conceptualization, data curation. B. Kazanowska: Conceptualization, data curation. C. Hutter: Conceptualization, data curation. S.M. Pfister: Conceptualization, data curation. D. Sturm: Conceptualization, data curation. D.T.W. Jones: Conceptualization, data curation. D. Orbach: Conceptualization, data curation. G. Pierron: Conceptualization, data curation. S. Raskin: Conceptualization, data curation. A. Drilon: Conceptualization, data curation. E.L. Diamond: Conceptualization, data curation. G. Harada: Conceptualization, data curation. M. Zapotocky: Conceptualization, data curation. J. Zamecnik: Conceptualization, data curation. L. Krskova: Conceptualization, data curation. B. Ellezam: Conceptualization, data curation. A.G. Weil: Conceptualization, data curation. D. Venne: Conceptualization, data curation. M. Barritault: Conceptualization, data curation. P. Leblond: Conceptualization, data curation. H. Coltin: Conceptualization, data curation. R. Hammad: Conceptualization, data curation. U. Tabori: Conceptualization, data curation. C. Hawkins: Conceptualization, data curation. J.R. Hansford: Conceptualization, data curation. D. Meyran: Conceptualization, data curation. C. Erker: Conceptualization, data curation. K. McFadden: Conceptualization, data curation. M. Sato: Conceptualization, data curation. N.G. Gottardo: Conceptualization, data curation. H. Dholaria: Conceptualization, data curation. D.S. Nørøxe: Conceptualization, data curation. H. Goto: Conceptualization, data curation. D.S. Ziegler: Conceptualization, data curation. F.Y. Lin: Conceptualization, data curation. D.W. Parsons: Conceptualization, data curation. H. Lindsay: Conceptualization, data curation. T.-T. Wong: Conceptualization, data curation. Y.-L. Liu: Conceptualization, data curation. K.-S. Wu: Conceptualization, data curation. A.T. Franson: Conceptualization, data curation. E. Hwang: Conceptualization, data curation. A. Aguilar-Bonilla: Conceptualization, data curation. S. Cheng: Conceptualization, data curation. C. Cacciotti: Conceptualization, data curation. M. Massimino: Conceptualization, data curation. E. Schiavello: Conceptualization, data curation. P. Wood: Conceptualization, data curation. L.M. Hoffman: Conceptualization, data curation. A. Cappellano: Conceptualization, data curation. A. Lassaletta: Conceptualization, data curation. A. Van Damme: Conceptualization, data curation. A. Llort: Conceptualization, data curation. N.U. Gerber: Conceptualization, data curation. M. Spalato Ceruso: Conceptualization, data curation. A.E. Bendel: Conceptualization, data curation. M. Skrypek: Conceptualization, data curation. D. Hamideh: Conceptualization, data curation. N. Mushtaq: Conceptualization, data curation. A. Walter: Conceptualization, data curation. N. Jabado: Conceptualization, data curation. A. Alsahlawi: Conceptualization, data curation. J.-P. Farmer: Conceptualization, data curation. C. Coleman: Conceptualization, data curation. S. Mueller: Conceptualization, data curation. C. Mazewski: Conceptualization, data curation. D. Aguilera: Conceptualization, data curation. N.J. Robison: Conceptualization, data curation. K. O’Halloran: Conceptualization, data curation. S. Abbou: Conceptualization, data curation. P. Berlanga: Conceptualization, data curation. B. Geoerger: Conceptualization, data curation. I. Øra: Conceptualization, data curation. C.L. Moertel: Conceptualization, data curation. E.D. Razis: Conceptualization, data curation. A. Vernadou: Conceptualization, data curation. F. Ducray: Conceptualization, data curation. C. Bronnimann: Conceptualization, data curation. R. Seizeur: Conceptualization, data curation. M. Clarke: Conceptualization, data curation. A.C. Resnick: Conceptualization, data curation. M. Alves: Conceptualization, data curation. C. Jones: Conceptualization, data curation. F. Doz: Conceptualization, data curation. T.W. Laetsch: Conceptualization, data curation. S. Perreault: Conceptualization, data curation, supervision, methodology, writing–original draft, writing–review and editing.

We gratefully acknowledge the invaluable support and expertise in data and statistical analysis provided by Mariève Cossette, MSc., and Marie-Claude Guertin, Ph.D., of the Montreal Health Innovations Coordinating Center. M. Zapotocky PRIMUS/19/MED/06 Charles University Grant Agency, Prague, Czech Republic; National Institute for Cancer Research (Program EXCELES, ID Project No. LX22NPO5102); European Union - Next Generation EU; Ministry of Health of the Czech Republic, grant nr (NU21-07-00419). D. Orbach is supported by a grant from SFCE-Imagine For Margo. A. Drilon National Cancer Institute/National Institutes of Health (P30CA008748), (1R01CA251591001A1), (1R01CA273224-01); Lungevity grants.